viewers. Before the end of the training, they selected three fellow workers whom they would prefer to have on their team. (These choices were actually taken into consideration when the teams were formed.) The votes were quite widely distributed, only three of the seventeen interviewers receiving no vote at all and the highest number of votes received by any interviewer was eight. The sociometric votes and those rating their fellow interviewers on ability, which were made at the same time, were positively related, but not very highly (rank order correlation = .39). Although there was a tendency for interviewers to have fellow workers of high ability on their team, this was not the only, or even the most important, consideration for choice.
The three ratings, poise, etc., in the screening, the general impression in the screening, and the sociometric rating can be taken as indications of the same traits which the tests measured, the ability to interact effectively with other people.
The coefficient of concordance32 for these five measures is .44 (significant at the one percent level). A combined ranking had low correlation with editing errors (.25), and a small negative correlation (-.17) with the supervisors' ratings. Effective dealing in personal relations does not seem to extend to dealing with superiors.
The second personality cluster, compliance, was based exclusively on one test, the Guilford-Martin Personnel Inventory.33 This test is designed to help supervisors to spot those workers who would be maladjusted and discontented and those who would fit into a work group. It is a questionnaire which is scored for three traits: "objectivity," "agreeableness," and "cooperativeness." It has previously been used successfully by research agencies for their interviewers.
The three scales had moderate correlations with each other, but not with any outside measure. The three correlations between scales were .28, .39, and .58.34 As the test had shown its usefulness before, the ranks of the three scores were averaged to obtain a separate cluster, which will be called compliance. It is correlated virtually zero with the number of errors in the first interviews and has a sizeable negative correlation (-.44) with the supervisors' ratings.
Of the two intermediate criteria, editing errors relates to intellectual abilities, but not to personality, and the supervisors' ratings did not relate to any of the three clusters. The question arises as to what criteria the supervisors use for their ratings.
A fourth cluster consists of the items which do relate to the supervisors' ratings. Two of these items were taken from the personal data form; these were age and experience. However, experience related negatively to the supervisors' rating and we shall take lack of experience as the item.
The other two items came from the questionnaire which the interviewers completed at the beginning of the training and which asked for their prediction of the respondent's reaction (cf. p. 29). In these the interviewers estimated the proportion of refusals which they would get and of respondents who would give reliable information. Optimism regarding these two questions is included in the cluster. These traits seem to define a person who is highly motivated. She is more likely to need the job (age), interviewing represents a distinct improvement over all previous jobs (lack of experience) and she pledges herself to make a painstaking effort (the ratings of expectancy). It is likely that a supervisor will find an interviewer of this kind easy to deal with. The four measures interrelate among each other less than was the case with the other clusters (W = .21, not significant), but this cluster is kept because of the importance of the supervisors' rating. The correlation of this cluster with the supervisor rating is .80. We shall call the cluster motivation.
The four clusters: intellectual ability; autonomy; compliance; and motivation are almost completely independent of each other, with the exception of a substantial, (but not significant) negative correlation between motivation and compliance. Interviewers may be high or low on any one of them without it being possible to predict how they will score on the other traits.
The task of the interviewer is to receive the information from the respondent and to transmit it for analysis. For this function the information has to be complete, accurate, and not influenced by the interviewer himself. The value of an interviewer's work, then, must be determined from the interviews he collects and final evaluation can only be made from all the interviews collected in the survey. Thus, the performance measures could only be determined after the fact and had no administrative significance in the conduct of the survey. They are, however, exceedingly important for subsequent work, for understanding patterns of interviewer performance, for evaluating the interviewer traits, and for establishing standards for interview quality.
The evaluation criteria used in this study were set up from all the interviews collected in the survey. It is conceivable that smaller numbers of interviews can be used____________________